[Tweet “Magic Beans for All! See it for yourself: #SQLServer Consolidation using #HyperV”]
You probably know the story “Jack and the Beanstalk” which tells of Jack, a young boy who trades a familiar solution – his cow, the family’s source of milk and income – for an unfamiliar solution that changes his life – some magic beans.
Fortunately for Jack, in this fairytale, the unfamiliar solution led to wealth, security, and fame. You might be hesitant to try something similar in your data center, but let me describe a modern “unfamiliar solution” to SQL Server workload consolidation using Hyper-V and the outcome it delivered. This solution:
- Increases by 3x the virtualized workloads your existing host can support
- Gives those 3x more workloads a 2x performance boost
- Reduces by 65% the total cost to consolidate 21 SQL Server workloads
- Delivers 79x ROI
Testing: SQL Server Consolidation using Hyper-V
SanDisk® and DB Best collaborated to compare three system configurations:
- A baseline system – using 24 x 300GB 15k RPM SAS hard drives.
- A “caching system” – the same system, but adding FlashSoft® caching software and one 1.3TB Fusion ioMemory™ PCIe flash application accelerator to hold cached data.
- An “all-flash system” – the baseline system with the hard drive storage replaced with two 6.4TB Fusion ioMemory cards.
Adi Cohen and Bob Bunker, both Senior Database Architects at DB Best, and Brian Walters, Senior Staff Applications Engineer at SanDisk’s Data Propulsion Lab, tested the workload configurations and found that:
The baseline system maxed out with
- 7 workloads
- 32,000 transactions/second (TPS)
- 68% CPU utilization
The caching system supported
- 18 workloads
- 67,000 TPS
- 75% CPU utilization
- With 21 workloads, this system still delivered 64,000 TPS at 69% CPU utilization
The all-flash system supported
- 25 workloads
- 85,000 TPS
- 89% CPU utilization
Do More, Save More
The detailed results are available in our whitepaper, but before I give you the link, let’s see what makes these unfamiliar solutions so compelling.
In summarizing the test results, we chose to focus on the consolidation of 21 workloads, because it allowed for convenient comparisons.
Using the baseline configuration with hard drives would require three hosts with each running 7 workloads, to support all 21 workloads; while both the caching and all-flash configurations each required only one host to support all 21 workloads.
The savings from needing only one host (instead of three) and licensing only 28 cores (instead of 84) to support the same 21 workloads, yielded:
- Caching system: $467,850 total system savings and a 79x ROI using FlashSoft caching software and one Fusion ioMemory card.
- All-flash system: $453,219 total system savings and a 22x ROI using only Fusion ioMemory.
A side benefit of the caching system is a 2x increase in the performance of those 21 workloads, and a 2.6x increase in performance with the all-flash system. Put 3x more workloads on these systems, and they still run faster – that’s definitely an unfamiliar solution!
Magic Beans for Everyone
These aren’t magic beans, they’re solutions that are shipping today. You may not get wealth, security, and fame like Jack did, but saving $450,000 might provide a little job security, and a modicum of fame. Not a bad deal.
The whitepaper Increase Hyper-V VM Density and Performance While Saving on Software License Cost provides a detailed treatment of the system configurations including detailed Bill of Materials, test procedure, and test results. It also offers guidance on when to choose each of the SanDisk solutions.
We’re thrilled to join DB Best in finding solutions to increase the efficiency of IT organizations as well as cloud and hosting providers. To find out more download the white paper here or get a custom assessment of your environment by DB Best to evaluate the savings and performance impact your organization will see.